Hello I'm

Talha Ejaz

As a System Engineer at JBT Corporation, I deploy AGVs, optimize navigation systems, and enhance performance with AI and ML. With a full-ride MS in Robotics and a Mechatronics degree, I offer strong expertise in system efficiency. My collaborative research across various fields and published work in Autonomous Navigation, Computer Vision, Renewable Energy, and Electric Vehicles highlight my innovative edge in robotics.

View Resume

Experience

At JBT Automated Systems, I worked as a Robotics System Installation Engineer, where I contributed to research and development initiatives focused on perception, sensor fusion, and control strategies for AGVs. I coordinated testing, installed AGVs and peripheral equipment, optimized mapping and navigation systems, documented system requirements, and provided guidance on system architecture and communication protocols to ensure effective implementation and operational efficiency.

At Georgia Tech, I worked as a Summer Research intern, focusing on developing Azure-based web applications with advanced GIS and data analysis functionalities, using tools like GeoPandas. My role involved optimizing performance and integrating DevOps practices, alongside leveraging AWS and Kubernetes for efficient deployment.

At Columbus State University, as a Teaching and Research Assistant, I mentored undergraduate students at Robotics Department provide academic support to faculty, focusing on teaching ROS and Linux. My responsibilities included overseeing students, guiding them in hardware and software concepts, organizing live coding sessions, and DSA coding exams for career development. I also conducted recitations for courses ENGR5238, and managed labs focused on 3D modeling, while mentoring students in these technical areas.

AI Engineer at Motiventive (remote), I led a team of 3 AI Engineers to develop emotion detection software for improved user engagement. Achieved a 15% processing efficiency boost by implementing precise, optimized image processing algorithms.

Trainee Engineer at Yunus Textile Mills Ltd., I collaborated with various stakeholders to meet project deadlines. I made data-driven decisions to control costs, presented weekly production reports to the HoD and directors, and offered strategic advice based on my analyses. Additionally, I was involved in troubleshooting and played a crucial role in the installation, erection, and commissioning of machinery for plant expansion projects.

Projects

Self Driving Car Using Deep Learning on Quanser Qcar

This project introduces a Deep Convolutional Neural Network (DCNN) architecture for autonomous navigation in robotics, enabling obstacle detection and avoidance. Mimicking human brain functions, the DCNN is trained to differentiate between static and dynamic obstacles, ensuring collision-free movement. The system, developed in Python 3.0, was tested using a Quanser Qcar equipped with an Nvidia Jetson TX2, demonstrating the network's effectiveness in real-world scenarios

GitHub Repository

Clearpath Jackal Frontier Exploration SLAM

Implemented Frontier-based exploration using LiDAR in ROS-Melodic Gazebo for the simulated environment for autonomous mapping of unknown territories. The independent mapping process involves localization, mapping, and exploration. With the emergence of unmanned aerial vehicles, there is now a need for autonomous exploration algorithms that work in tandem with simultaneous localization and mapping (SLAM).

GitHub Repository

Indoor Localization with IMU-based Systems using Kalman Filters (KF, EKF, UKF)

The project evaluates the performance of three different Kalman filtering techniques KF, EKF, and UKF to enhance position accuracy within an IMU-based indoor localization system. Each technique's effectiveness in refining position estimates from raw sensor data will be rigorously tested in two distinct campus buildings using pyhton.

GitHub Repository

Space Invaders Game using Deep Reinforcement Learning

Implement a deep learning model to successfully learn control policies directly from high-dimensional sensory inputs using reinforcement learning. We chose to work with RAM Monitor. In this case, environmental monitoring is provided by the state of the RAM, that is, a set of 128 values. Each value is a byte (an integer from 0 to 255). RAM is a compact representation of state compared to algorithms that display the image as a state. In this environment The reward in the Space Invaders gym environment is the result in the given time step.

GitHub Repository

Semantic Segmentation

Semantic Segmentation is a special form of image segmentation that deals with detecting of objects. The model is trained in COCO dataset and by using PixelLib we test the model at Columbus State University parking area where the camera is placed in the dashboard of the car.

GitHub Repository